OWT MULTICHANNEL SURE-LET SOFTWARE FOR MATLAB
Florian Luisier
Biomedical Imaging group
Swiss Federal Institute of Technology Lausanne (EPFL), Switzerland.
2006-2010
This software implements the algorithm described in:
[1] F. Luisier, T. Blu, "SURE-LET Multichannel Image Denoising: Interscale Orthonormal Wavelet Thresholding," IEEE Transactions on Image Processing, vol. 17, no. 4, pp. 482-492, April 2008.
[2] F. Luisier, "The SURE-LET Approach to Image Denoising," Swiss Federal Institute of Technology Lausanne, EPFL Thesis no. 4566 (2010), 232 p., January 8, 2010.
A. TEST PROGRAM
DENOISING_DEMO: Denoising demonstration based on the multichannel SURE-LET principle applied to interscale orthonormal wavelet thresholding. To run this script, just type 'denoising_demo' in your Matlab Command Window.
Note that the core of this demo is the denoising step that is performed by the command:
output = OWT_MC_SURELET_denoise(input,wtype,R);
B. FFT-BASED WAVELET TRANSFORM FUNCTIONS
FFT_WAVEFILTERS: Computes the frequency responses of the wavelet analysis and synthesis filters (lowpass and highpass) for various types of filters.
FFT_GDC_FILTER: Computes the frequency response of the Group Delay Compensation (GDC) filter associated to a particular orthonormal wavelet filter.
FFT_WAVEDEC: Performs a FFT-based computation of the discrete real wavelet transform of a 2D signal of size [Nx,Ny] to a given depth [Jx,Jy]. At least one dimension of the signal must be even.
FFT_WAVEREC: Performs a FFT-based computation of the inverse discrete wavelet transform.
C. DENOISING FUNCTIONS
OWT_MC_SURELET_DENOISE: Removes additive white Gaussian noise using the multichannel interscale SURE-LET principle in the framework of an orthonormal wavelet transform (OWT) only. The wavelet transform is included in the process.
FCN_MIN_MC_SURE: Removes additive white Gaussian noise inside a given wavelet subband by minimizing SURE, as described in [1,2].
FCN_MC_DENOISE: Removes additive white Gaussian noise using the multichannel interscale SURE-LET principle.
D. OTHER AUXILIARY ROUTINES
AUX_GAUSSIAN_SMOOTHING: Applies a normalized 2D Gaussian kernel on a 2D signal.
AUX_DYADIC_MAX_SCALES: Computes the maximum number of dyadic scales.
AUX_NOISE_ESTIM: Estimates the standard deviation of the additive white Gaussian noise, using a robust eigenfilter procedure.
AUX_NUM_OF_ITERS: Computes the most suitable number of iterations to be performed in the SURE-LET algorithm.
AUX_STACKREAD: Reads a stack of images and converts it to a Matlab 3D double matrix.
The folder also contains some standard test images (gray-level, color and multichannel).
**********************************************************
Please, report any bugs, comments or suggestions to:
florian.luisier@a3.epfl.ch
**********************************************************